Skip to content

PaperPilot v1.5.1

Choose a tag to compare

@CHB-learner CHB-learner released this 15 May 09:56
· 25 commits to main since this release

Fixed

  • Fixed non-RNA / non-AI biomedical topics being over-screened to zero papers.
  • Fixed query understanding prompts that could silently narrow broad biomedical requests into AI-only protocols.
  • Replaced the RNA-specific screening gate with topic-adaptive relevance scoring while preserving RNA inverse-folding exclusions.
  • Fixed the interactive welcome banner to show the package version dynamically instead of stale v1.3.

Added

  • Added a formal 30-paper minimum policy for generated literature reports.
  • Added --min-report-papers and --no-obsidian-wiki CLI options.
  • Added report_selection.json and shortfall.json diagnostics for minimum-report enforcement.
  • Added default Obsidian Wiki export under obsidian_wiki/ with paper, method, topic, claim, manifest, and lint files.

Changed

  • Final report selection is now core-first, then code-filter fallback, then adjacent-paper fill when needed.
  • Quality gate metrics now record minimum-report counts, fill counts, and code-filter fallback counts.
  • README, Chinese README, local usage docs, and project homepage now document the 30-paper policy and Obsidian Wiki output.

Added

  • Release infrastructure refinement: scripts/release_everywhere.sh now supports end-to-end local validation + dry-run with clean push/PyPI gating.
  • CLI/docs alignment clean-up: release helper examples now use version placeholders and avoid stale manual version pinning in docs.
  • Publishing workflow polish: stable dry-run behavior and deterministic changelog extraction for GitHub release notes.

Changed

  • Updated version tracking to keep pyproject.toml and literature_agent/__init__.py in sync during one-command release flows.
  • Improved release safety: skip-push / no-pypi / dry-run semantics are now consistently applied across publishing stages.

Added

  • Added multi-source registry expansion with papers.cool and deepxiv integration.
  • Added citation-numbered canonical reporting flow with consistent zh/en/html/pdf outputs and evidence mapping.
  • Added stronger workflow observability: richer run artifacts, source diagnostics, evidence ledger flow and doctor checks.
  • Added prompt registry improvements and synthesis depth enhancements (RQ framing, method taxonomy, claims/evidence map, contradiction tracking).
  • Added default ~/.paperpilot/config.json initialization behavior and optional source API slots in config templates.
  • Added GitHub homepage content under docs/ for project promotion.

Changed

  • Upgraded report synthesis pipeline to produce deeper review-style text (background, methods, evidence, trends, gaps).
  • Improved paper normalization and ranking robustness for mixed-type metadata.
  • Fixed citation manifest alignment and report table rendering stability.

Fixed

  • Fixed evidence ledger generation edge cases when claim IDs are absent.
  • Fixed workflow manifest version drift by binding to package version.
  • Fixed rank/synthesis pipeline compatibility with non-string metadata inputs.
  • Added source management commands and interactive /sources command.
  • Added doctor workflow and richer CLI presentation.